As for this species particularly, the answer is simple. There is no reliable evidence that of Mecidea prolixa is present in the New World. There is no reason to consider any such identification reliable.
But what about the broader situation? iNaturalist has a huge number of obviously erroneous identifications. Other Heteroptera that immediately come to mind (North American species on other continents) include:
As with Mecidea, for the majority of these observations, identification to the species level is technically impossible due to the difficulty of distinguishing the species.
What can we (the identifiers) do about this? We could thoroughly check and “correct” every entry to the genus level - something I actually do quite often. But does this solve the problem? I am not sure. There are few of us, and our time is limited. While I am busy doing this, truly interesting finds elsewhere are languishing, waiting to be identified. If I don’t do it, the site becomes filled with errors that degrade the CV model. It is, in fact, a difficult choice. I have simply resigned myself to the “European Orius insidiosus” issue. This species is frequently used as a biological control agent in greenhouses and could easily be introduced into Europe, yet there is currently no reliable evidence to confirm this.
Is a “batch fix” a good idea? I’m not so sure about that either. We might overlook the real cases of species introductions. Besides, “identification based solely on distribution” (rather than on the specimen’s features) strikes me as a questionable approach.
Sometimes it seems to me that these identification errors are simply the inevitable “price of freedom.” We have this unique dataset precisely because it is so easy and accessible to add and modify data. And inevitably, some of that data will turn out to be inaccurate or erroneous, for that very reason. “It’s not a bug; it’s a feature”.
But I’ll still be “checking and correcting” the obs from time to time.
I think it should be possible to program the CV so it is FORBIDDEN to suggest species for certain genera, tribes, complexes. We could still write in a species ID if we want to, but the CV wouldn’t even offer the option. (Process for forbidding should involve a flag and discussion.)
Another option is addressing this problem as eBird does. If you try to report a bird on eBird at a highly unlikely place or time or in a remarkably large number, eBird flags the report as unusual and asks the reporter to confirm that they are familiar with the species in question and are certain that they saw the species and number that they are reporting. If the reporter confirms their report, then eBird notifies an experienced reviewer to double-check the report, and until it has been double-checked, the flagged report is kept off of the species map. If iNat had a similar system, flagging records with extremely high geoscores or from an unprecedented country or continent, and asking for confirmation from the observer similarly to a record submitted without an ID or location, then categorizing the observation as “Casual” until a curator double-checks it, that would hugely reduce the frequency of problems like this.
Anyway, it looks like this specific problem (see what I did there ) is fixed for now! I guess, while I was sleeping, Sam and Brad bumped the (100+?) US records of M. prolixa up to genus. All that was left for me to do this morning was the same for the 30-ish records from Mexico and the Caribbean. The map for M. prolixa now zooms right in to southern Africa, as it should. Will try to return to the species in a few weeks to mop up new ones, and hopefully the stream of errors will ultimately dry up and blow away.
And maybe, since this took less than 24 hours, some of us can now help Konstantin with the European Orius insidiosus mess that he mentions, which doesn’t really look so much bigger than this one did, about 180 records involved, only half a dozen research-grade.
I don’t think it’s hopeless. I’ve cleaned up several smaller messes, and many of them seem to stay clean afterwards. For instance, a few months ago I noticed a bunch of Vespula acadia records from Europe. After confirming that the species had never been certifiably detected outside of North America, I spam-corrected all of the European records. As of today, no new errors of that type have popped up. I may have caught that one before enough errors accumulated to impact the AI model. Will keep on plugging away…
Hi Patrick! The Sinea issue is much more prickly (hehe, there I go again) because the two species overlap so broadly in distribution. The M. prolixa and Orius insidiosus situations are easier to deal with because they involve observations not just outside the species’ known range, but on entirely different continents, so the questionable ones can be fairly quickly identified, scrutinized, and if necessary corrected. Fixing Sinea will be much more tedious, especially since some people actually do get photos of adults that show the pronotum in detail, so that minority really can be identified to species, and there’s no quick and easy way to separate those from the observations that don’t show enough detail. Plus, most people don’t always use the life stage annotations which would allow quickly separating the unidentifiable nymphs from the occasionally identifiable adults. I think the only way through that tangle is what the video-gamers call “grinding”… “The Sorcerer’s Apprentice problem” is a great name for this syndrome, unless that pop-culture reference becomes too anachronistic to be widely recognizable except by us old people
I didn’t know that there was a name for that procedure, but the “Geomodel Anomaly Detector” is one of my favorite techniques for procrastinating household chores and other less pleasant duties!
Here to agree with Brad but also provide hope; I encountered a rampant CV issue with Maladeraformosae flooding out another invasive “june bug”, Nipponoserica peregrina, in addition to native Serica species. After many hours sorting out the Nipponoserica observations, I eventually noticed it start showing up in the CV suggestions, significantly cutting back the misidentifications in most regions.
Since this thread has become a general location for griping about CV-(mis)guided massive ID mistakes (“error-clouds”), I’ll chime in with a particular situation which requires my daily attention. Here’s my biogeographical conundrum:
The Two-banded Petrophila is common and widespread across eastern North America, including Texas, and is readily IDed (by humans and CV) over 90% of its range with 3,800+ RG observations. However, the Texas near-endemic Capps’ Petrophila overlaps the Texas distribution of Two-banded almost completely and has just 560 RG’s. (Two-banded is thus 7X more abundant than Capps’ overall and about 2-1/2X as abundant in Texas). Capps’ has an identical forewing pattern as Two-banded and can only be distinguished with a good view of the hindwings–only evident in a minority of Texas observations. CV hasn’t “learned” this nuance and blithely continues to offer Two-banded Petrophila as an ID for every Texas observation of this species pair including any/all observations lacking a view of the hindwings. I have accepted the task of backing off “no-hind-wing views” of this species pair in Texas to genus level (annoying to both OPs and me). During/after the City Nature Challenge, National Moth Week, and other organized bioblitz events, I end up spending as much as an hour a day correcting just this one species ID dilemma. It’s a burden I bear.
Another issue with Sinea is that there is another widespread species in the East, the (fairly) recently described Sinea incognita, said to be widespread in the Southeast to the Midwest. Usually has short wings. I’ve not seen it. Maybe I’ve mistaken for a nymph. I have detailed photos of several specimens of Sinea from DRM. They appear to be S. diadema, but a couple I cannot really tell, even with detailed photos. I also wonder if there might be some sexual dimorphism, but I see no info on that. It’s the sort of thing were it seems one just has to be comparing specimens in a museum. Seems like barcoding might help resolve some issues, but I don’t see that anyone has done it systematically.
As you noted, a big issue, still, on iNaturalist is that people are identifying nymphs to species, which seems not possible, mostly. (There is one paper where a couple of species were reared and each instar examined. Differences were discernible with a dissecting microscope.) Also, people were calling any nymph “spinipes”, but all species of nymphs are spiny. Oof. I feel like the whole group is just not identifiable beyond genus from the typical image taken in the field.
I had time to conduct a purge, and South America is now cleared out of Zelus luridus records! Well, except for one; it was ID’d by an entomologist from Columbia who says on his profile that he specializes in hemipteran diversity, and who acknowledges in his ID that the species has never been reported from that location before. The observation was made in the Galapagos, a popular enough tourist destination for a hitchhiker from North America to be all to believable, and isolated enough that it appears to lack any other species easily confused with Z. luridus. I did question the guy on why he selected that particular species rather than any of the similar-looking ones known from the South American mainland, like Atopozelus (which was iNat’s top suggestion for nearly all of the observations from the continent that had been ID’d as Z. luridus) Will see if and how he responds…
So, the next time I have the urge to purge, maybe I can take on Orius insidiosus, unless somebody else beats me to it!
Thank you very much for your efforts! Although I’m still not sure if this is the right approach. One of the most important things that’s only possible thanks to the community, it seems to me, is the opportunity to see “strange” findings. Like this one: https://www.inaturalist.org/observations/108659691. I’m afraid that a “batch downgrade of identifications” might make them less visible.
) is fixed for now! I guess, while I was sleeping, Sam and Brad bumped the (100+?) US records of M. prolixa up to genus. All that was left for me to do this morning was the same for the 30-ish records from Mexico and the Caribbean. The map for M. prolixa now zooms right in to southern Africa, as it should. Will try to return to the species in a few weeks to mop up new ones, and hopefully the stream of errors will ultimately dry up and blow away.